1. Field of the Invention
This invention relates to electronic systems, and more particularly to electrical interconnecting apparatus having continuous planar conductors.
2. Description of the Related Art
Electronic systems typically employ several different types of electrical interconnecting apparatus having planar layers of electrically conductive material (i.e., planar conductors) separated by dielectric layers. A portion of the conductive layers may be patterned to form electrically conductive signal lines or xe2x80x9ctracesxe2x80x9d. Conductive traces in different layers (i.e., on different levels) are typically connected using contact structures formed in openings in the dielectric layers (i.e., vias). For example, integrated circuits typically have several layers of conductive traces which interconnect electronic devices formed upon and within a semiconductor substrate. Each layer is separated from adjacent layers by dielectric layers. Within a semiconductor device package, several layers of conductive traces separated by dielectric layers may be used to electrically connect bonding pads of an integrated circuit to terminals (e.g., pins or leads) of the device package. Printed circuit boards (PCBs) also typically have several layers of conductive traces separated by dielectric layers. The conductive traces are used to electrically interconnect terminals of electronic devices mounted upon the PCB.
Signals in digital electronic systems typically carry information by alternating between two voltage levels (i.e., a low voltage level and a high voltage level). A digital signal cannot transition instantaneously from the low voltage level to the high voltage level, or vice versa. The finite amount of time during which a digital signal transitions from the low voltage level to the high voltage level is called the rise time of the signal. Similarly, the finite amount of time during which a digital signal transitions from the high voltage level to the low voltage level is called the fall time of the signal.
Digital electronic systems are continually being produced which operate at higher signal frequencies (i.e., higher speeds). In order for the digital signals within such systems to remain stable for appreciable periods of time between transitions, the rise and fall times of the signals must decrease as signal frequencies increase. This decrease in signal transition times (i.e., rise and fall times) creates several problems within digital electronic systems, including signal degradation due to reflections, power supply xe2x80x9cdroopxe2x80x9d, ground xe2x80x9cbouncexe2x80x9d, and increased electromagnetic emissions. It is desirable that the digital signals are transmitted and received within accepted tolerances.
A signal launched from a source end of a conductive trace suffers degradation when a portion of the signal reflected from a load end of the trace arrives at the source end after the transition is complete (i.e., after the rise time or fall time of the signal). A portion of the signal is reflected back from the load end of the trace when the input impedance of the load does not match the characteristic impedance of the trace. When the length of a conductive trace is greater than the rise time divided by three, the effects of reflections upon signal integrity (i.e., transmission line effects) should be considered. If necessary, steps should be taken to minimize the degradations of signals conveyed upon the trace due to reflections. The act of altering impedances at the source or load ends of the trace in order to reduce signal reflections is referred to as xe2x80x9cterminatingxe2x80x9d the trace. For example, the input impedance of the load may be altered to match the characteristic impedance of the trace in order to prevent signal reflection. As the transition time (i.e., the rise or fall time) of the signal decreases, so does the length of trace which must be terminated in order to reduce signal degradation.
A digital signal alternating between the high and low voltage levels includes contributions from a fundamental sinusoidal frequency (i.e., a first harmonic) and integer multiples of the first harmonic. As the rise and fall times of a digital signal decrease, the magnitudes of a greater number of the integer multiples of the first harmonic become significant. As a general rule, the frequency content of a digital signal extends to a frequency equal to the reciprocal of xcfx80 times the transition time (i.e., rise or fall time) of the signal. For example, a digital signal with a 1 nanosecond transition time has a frequency content extending up to about 318 MHz.
All conductors have a certain amount of inductance. The voltage across the inductance of a conductor is directly proportional to the rate of change of current through the conductor. At the high frequencies present in conductors carrying digital signals having short transition times, a significant voltage drop occurs across a conductor having even a small inductance. A power supply conductor connects one terminal of an electrical power supply to a power supply terminal of a device, and a ground conductor connects a ground terminal of the power supply to a ground terminal of the device. When the device generates a digital signal having short transition times, high frequency transient load currents flow in the power supply and ground conductors. Power supply droop is the term used to describe the decrease in voltage at the power supply terminal of the device due to the flow of transient load current through the inductance of the power supply conductor. Similarly, ground bounce is the term used to describe the increase in voltage at the ground terminal of the device due to the flow of transient load current through the inductance of the ground conductor. When the device generates several digital signals having short transition times simultaneously, the power supply droop and ground bounce effects are additive. Sufficient power supply droop and ground bounce can cause the device to fail to function correctly.
Power supply droop is commonly reduced by arranging power supply conductors to form a crisscross network of intersecting power supply conductors (i.e., a power supply grid). Such a grid network has a lower inductance, hence power supply droop is reduced. A continuous power supply plane may also be provided which has an even lower inductance than a grid network. Placing a xe2x80x9cbypassxe2x80x9d capacitor near the power supply terminal of the device is also used to reduce power supply droop. The bypass capacitor supplies a substantial amount of the transient load current, thereby reducing the amount of transient load current flowing through the power supply conductor. Ground bounce is reduced by using a low inductance ground conductor grid network, or a continuous ground plane having an even lower amount of inductance. Power supply and ground grids or planes are commonly placed in close proximity to one another in order to further reduce the inductances of the grids or planes.
Electromagnetic interference (EMI) is the term used to describe unwanted interference energies either conducted as currents or radiated as electromagnetic fields. High frequency components present within circuits producing digital signals having short transition times may be coupled into nearby electronic systems (e.g., radio and television circuits), disrupting proper operation of these systems. The United States Federal Communication Commission has established upper limits for the amounts of EMI products for sale in the United States may generate.
Signal circuits form current loops which radiate magnetic fields in a differential mode. Differential mode EMI is usually reduced by reducing the areas proscribed by the circuits and the magnitudes of the signal currents. Impedances of power and ground conductors create voltage drops along the conductors, causing the conductors to radiate electric fields in a common mode. Common mode EMI is typically reduced by reducing the impedances of the power and ground conductors. Reducing the impedances of the power and ground conductors thus reduces EMI as well as power supply droop and ground bounce.
Within the wide frequency range present within electronic systems with digital signals having short transition times, the electrical impedance between any two parallel conductive planes (e.g., adjacent power and ground planes) may vary widely. The parallel conductive planes may exhibit multiple electrical resonances, resulting in alternating high and low impedance values. Parallel conductive planes tend to radiate a significant amount of differential mode EMI at their boundaries (i.e., from their edges). The magnitude of differential mode EMI radiated from the edges of the parallel conductive planes varies with frequency and is directly proportional to the electrical impedance between the planes.
FIG. 1 is a perspective view of a pair of 10 in.xc3x9710 in. square conductive planes 110 and 120 separated by a fiberglass-epoxy composite dielectric layer. Each conductive plane is made of copper and is 0.0014 in. thick. The fiberglass-epoxy composite layer separating the planes has a dielectric constant of 4.0 and is 0.004 in. thick. If a 1 ampere constant current is supplied between the centers of the rectangular planes, with a varying frequency of the current, the magnitude of the steady state voltage between the centers of the rectangular planes can be determined 130.
The electrical impedance between the parallel conductive planes of FIG. 1 varies widely at frequencies above about 200 MHz. The parallel conductive planes exhibit multiple electrical resonances at frequencies between 100 MHz and 1 GHz and above, resulting in alternating high and low impedance values. The parallel conductive planes of FIG. 1 would also radiate substantial amounts of EMI at frequencies where the electrical impedance between the planes anywhere near their peripheries is high.
The above problems are currently solved in different ways at different frequency ranges. At low frequency, the power supply uses a negative feedback loop to reduce fluctuations. At higher frequencies, large value bypass (i.e. decoupling) capacitors are placed near devices. At the highest frequencies, up to about 200-300 MHz, very small bypass capacitors are placed very close to devices in an attempt to reduce their parasitic inductance, and thus high frequency impedance, to a minimum value. By Nov. 2, 1994, the practical upper limit remained around 200-300 MHz as shown by Smith [Decoupling Capacitor Calculations for CMOS Circuits; pp. 101-105 in Proceedings of 3rd Topical Meeting on Electrical Performance of Electronic Packaging of the Institute of Electrical and Electronics Engineers, Inc.].
The power distribution system was modeled as shown in FIG. 2. A switching power supply 210 supplies current and voltage to a CMOS chip load 220. In parallel with the power supply 210 and the load 220 are decoupling capacitors 215 and the PCB 225 itself, with its own capacitance. Smith [1994] teaches that decoupling capacitors are only necessary up to 200-300 MHz, as the target impedances are rarely exceeded above that frequency. This upper limit changes over time as the clock frequencies increase and the allowable voltage ripple decreases. Determining the proper values for decoupling capacitors and the optimum number of each has been a xe2x80x9ctrial and errorxe2x80x9d process, which relies on the experience of the designer. There are no known straightforward rules for choosing decoupling capacitors for all frequency ranges.
It would thus be desirable to have a method for designing the power distribution system and determining the desired decoupling components for stabilizing the electrical impedance in the power distribution system. The method is preferably automatable using a computer system to perform calculations.
The problems outlined above are in large part solved by a method for determining the desired decoupling components for stabilizing the electrical impedance in the power distribution system of an electrical interconnecting apparatus including a pair of parallel planar conductors separated by a dielectric layer. The electrical interconnecting apparatus may be, for example, a PCB, a semiconductor device package substrate, or an integrated circuit substrate. The present method includes creating a model of the power distribution system based upon an Mxc3x97N grid for both the power plane and the ground plane. The model is preferably based upon a fixed grid which adapts automatically to the actual physical dimensions of the electrical interconnecting apparatus. The model preferably also calculates the system response to chosen decoupling components in both single node and Mxc3x97N node versions.
The model receives input from a user and from a database of various physical and/or electrical characteristics for a plurality of decoupling components. Various characteristics of interest include physical dimensions, type and thickness of dielectric, method and materials of manufacture, capacitance, mounted inductance, and ESR. The desired characteristics are preferably saved in a database for corrections, additions, deletions, and updates.
In one embodiment, the model of the power distribution system is in a form of a plane including two dimensional distributed transmission lines. The model of the power distribution system may comprise a plurality of the following: one or more physical dimensions of the power plane, one or more physical dimensions of the ground plane, physical separation distance of the power plane and the ground plane, composition of a dielectric separating the power plane and the ground plane, one or more active device characteristics, one or more power supply characteristics, and one or more of the decoupling components. In one embodiment, M and N have a uniform value. In various embodiments, the active components act as current sources or sinks, and may include processors, memories, application specific integrated circuits (ASICs), logic ICs, or any device which converts electrical energy into information. Preferably, a total sum of all values of the current sources in the model is scaled to equal one ampere.
In one embodiment, the model of the power distribution system is operable for determining the decoupling components for a frequency range above approximately a lowest board resonance frequency. In another embodiment, the model of the power distribution system is operable for determining the decoupling components for a frequency range above a highest resonance frequency from all resonance frequencies of the decoupling components.
The method preferably includes determining a target impedance for the power distribution system at a desired frequency or over a desired frequency range. The target impedance is preferably determined based upon such factors as power supply voltage, total current consumption, and allowable voltage ripple in the power distribution system. Preferably, determining the target impedance for the power distribution system comprises the quotient of power supply voltage multiplied by allowable voltage ripple divided by total current.
The frequency range may start at dc and rise to several GHz. In one embodiment, the model of the power distribution system is operable for determining the decoupling components for a frequency range above approximately a lowest board resonance frequency. In another embodiment, the model of the power distribution system is operable for determining the decoupling components for a frequency range above a highest resonance frequency from all resonance frequencies of the decoupling components.
The method preferably selects one or more decoupling components from a plurality of possible decoupling components. Preferably, the decoupling components are capacitors with an approximate mounted inductance and an ESR. In one embodiment, a range of the values of the capacitors is chosen such that a superposition of impedance profiles provide an impedance at or below the target impedance for the power distribution system over the frequency range of interest. In one embodiment, the impedance profiles of the plurality of possible decoupling components are compared to resonance frequencies for the power distribution system. The decoupling components have resonance frequencies which substantially correspond to the resonance frequencies of the power distribution system in the frequency range of interest.
The method preferably determines a number for each of the one or more decoupling components chosen to be included as part of the power distribution model. In one embodiment, the number of the various decoupling components is chosen based upon the frequency range of interest and the impedance profiles of a plurality of possible decoupling components. In another embodiment, the number of a particular one of the decoupling components is chosen to have approximately equal value of a next larger integer of the quotient obtained from dividing the ESR for the particular decoupling components by the target impedance for the power distribution system. In still another embodiment, the number of particular decoupling components has approximately equal value of impedance to the target impedance for the power distribution system when the number of the particular decoupling components are placed in parallel. In one embodiment, determining the number for the each of the decoupling components occurs before effectuating the model of the power distribution system to determine the transfer impedance values as the function of frequency at the one or more specific locations.
The method preferably places one or more current sources in the model of the power distribution system at one or more spatial locations corresponding to one or more locations of active components. The method also preferably places the decoupling components in the model of the power distribution system at nodal points that couple the Mxc3x97N grid for the power plane and the corresponding Mxc3x97N grid for the ground plane. In one embodiment, the method places a power supply in the model of the power distribution system at a fixed location on the power plane. The power supply is preferably comprised in the model as one or more pole frequencies, one or more zero frequencies, and one or more resistances.
The method preferably selects one or more specific locations in the model of the power distribution system to calculate transfer impedance values as a function of frequency. The method preferably effectuates the model of the power distribution system to determine the transfer impedance values as the function of frequency at the one or more specific locations previously chosen. The method then preferably compares the transfer impedance values as the function of frequency at the one or more specific locations to the target impedance for the power distribution system. Preferably, the method determines a bill of goods for the power distribution system based upon the results of effectuating the model.
In various embodiments, the method for determining decoupling components for a power distribution system includes determining a preferred or optimum number of decoupling components for a power distribution system. A preferred method for determining a number of decoupling components for a power distribution system is also disclosed. For a given frequency or frequency range, the method for determining a number of decoupling components for a power distribution system comprises selecting a particular one of the decoupling components based upon a mounted inductance of each of the decoupling components. The mounted inductance comprises an indication of a resonance frequency of that particular one of the decoupling components. The method also compares an individual decoupling component impedance of each of the decoupling components to the target impedance. The method then selects the number of the particular one of the decoupling components which provides a total impedance at or below the target impedance at the given frequency or frequency range.
In one embodiment, if the impedance of the particular decoupling component is greater than the target impedance, then the method calculates the desired number of the particular decoupling components in a parallel configuration. In embodiments that determine a number of each of a plurality of decoupling components for a power distribution system for a given frequency range, a plurality of decoupling components are chosen as necessary to provide a total impedance at or below the target impedance for the given frequency range.
In various embodiments, the method for determining decoupling components for a power distribution system includes determining placement information for preferred or optimum number of decoupling components for a power distribution system. A preferred method for determining placement of one or more decoupling components in a power distribution system is also given. In one embodiment, each of the one or more decoupling components includes a respective resonance frequency and a respective equivalent series resistance at the respective resonance frequency. The power distribution system includes a target impedance, and the electrical interconnecting apparatus has at least a first dimension. The method determines one or more board resonance frequencies. A first board frequency corresponds to the first dimension. The method also selects one or more first decoupling components from a plurality of possible decoupling components such that the first decoupling components have their respective resonance frequency at approximately the first board resonance frequency. The method then places the first decoupling components on a location of the electrical interconnecting apparatus corresponding to the first dimension. Additional dimensions of the electrical interconnecting apparatus may also require their own decoupling components.
In the embodiment where the electrical interconnecting apparatus has approximately rectangular dimensions, the first dimension is preferably an effective length and the second dimension is preferably an effective width. The preferred location for placing the decoupling component for the first dimension comprises a first edge on the effective length, while the preferred location for placing the decoupling component for the second dimension comprises a second edge on the effective width.
In one embodiment, when the electrical interconnecting apparatus has at least one location for at least a first active device, the method further comprises placing one or more second decoupling components on the electrical interconnecting apparatus at the at least one location for at least the first active device. Additional decoupling components are also placed on the electrical interconnecting apparatus as needed for additional active devices. The preferred location for placing decoupling components for active devices is at or near the active devices.
In one embodiment, the method includes selecting the decoupling components from a plurality of possible decoupling components such that the decoupling components have the respective resonance frequency at approximately the first operating frequency of the active device. Additional decoupling components may be selected and placed based upon the harmonics of the operating frequency, as desired.